### Summary Split our `Tensor` into three buckets: * 11 tensors archetypes (`TensorU8`, `TensorU16`, `TensorU32`, …) backed by arrow-typed buffer (store f32 tensor as as f32 array) * `Image` + `DepthImage` + `SegmentatioImage` archetypes: backed by `PixelBuffer` + `PixelFormat` + `Size` (“straight from a sensor, straight to the GPU”) * `ImageEncoded` archetype, backed by ByteBlob + MediaType (JPEG, PNG, …) ### Roadmap * https://github.com/rerun-io/rerun/issues/3803 * https://github.com/rerun-io/rerun/pull/6882 * https://github.com/rerun-io/rerun/pull/6874 * https://github.com/rerun-io/rerun/pull/6883 * https://github.com/rerun-io/rerun/pull/6884 * https://github.com/rerun-io/rerun/issues/6386 * https://github.com/rerun-io/rerun/pull/6915 * https://github.com/rerun-io/rerun/pull/6928 * https://github.com/rerun-io/rerun/pull/6942 * https://github.com/rerun-io/rerun/issues/7050 * [ ] Figure out what to do with `BarChart`:s `TensorData` * https://github.com/rerun-io/rerun/issues/6832 * https://github.com/rerun-io/rerun/issues/6830 ### Other issues and links * https://github.com/rerun-io/rerun/issues/3804 * https://github.com/rerun-io/rerun/issues/6388#issuecomment-2130162089 * https://github.com/rerun-io/rerun/issues/6388#issuecomment-2133987252 ### Part of * #6388 * #6819
Summary
Split our
Tensorinto three buckets:TensorU8,TensorU16,TensorU32, …) backed by arrow-typed buffer (store f32 tensor as as f32 array)Image+DepthImage+SegmentatioImagearchetypes: backed byPixelBuffer+PixelFormat+Size(“straight from a sensor, straight to the GPU”)ImageEncodedarchetype, backed by ByteBlob + MediaType (JPEG, PNG, …)Roadmap
TensorBuffer::JPEGwithImageEncodedarchetype #3803ImageEncodedtoImageEncodedHelper#6882archetypes.ImageEncodedwith PNG and JPEG support #6874ImageChromaDownsampled#6883TensorBuffer::JPEG,DecodedTensor,TensorDecodeCache#6884DepthImagearchetype #6915SegmentationImageto the new image archetype style #6928Imagearchetype #6942BarChart:sTensorDataTensorDimensioninto shape and names #6830Other issues and links
Part of